🥑 Meta pushes AI model 'Avocado' rollout to May or later, NYT reports ↗
Meta’s next big AI model, code-named Avocado, has slipped yet again. It was meant to help prove Meta could keep pace at the very top tier, but internal performance apparently has not been strong enough, which is not a small problem when your rivals are OpenAI, Google and Anthropic.
The more peculiar part is that Meta still sounds fully committed, merely less punctual. There is also talk that it considered licensing Google’s Gemini technology as a stopgap for some products - which would amount to a stark admission that building frontier AI and shipping it at scale are two very different feats.
⚖️ Artificial Intelligencer: Anthropic's delicate dance with the Pentagon ↗
Anthropic’s clash with the Pentagon kept widening, and the stakes now appear immense. The company is contesting a designation that blocks its technology from certain military-related use, while arguing that the fallout could hit contracts, partnerships and a large share of revenue that was expected to materialise.
It is one of those stories that looks legal on the surface and existential underneath. Anthropic has tried to hold a line on military applications, but that position is now colliding with government power, enterprise anxiety, and the broader question of who gets to build AI for national security - and on what terms.
🎓 New ways to learn math and science in ChatGPT ↗
OpenAI rolled out dynamic visual explanations in ChatGPT for more than 70 core math and science concepts. Instead of merely describing formulas and relationships, ChatGPT can now show them changing in real time, which may sound obvious in hindsight, though it is not.
This nudges ChatGPT a little further into the “learning tool, not just answer machine” lane. The move fits with OpenAI’s broader attempt to make the product more interactive and less fixed - more whiteboard, less encyclopedia, or so it seems.
🗺️ How we’re reimagining Maps with Gemini ↗
Google brought more Gemini into Maps with a conversational feature called Ask Maps, alongside a revamped navigation experience. The pitch is simple enough: ask complex real-world questions, get personalised place suggestions on a map, then move straight into planning without bouncing between apps.
The navigation side sounds just as ambitious, with more immersive visuals and clearer route guidance. It is Google leaning into a familiar idea - search, but spatial and conversational - and trying to make Maps feel less like a utility, more like a local guide who knows the shortcuts.
🏢 Google names London office 'Platform 37' in a nod to railway neighbour, AI 'Go' match ↗
Google gave its new London headquarters the name Platform 37, tying it both to King’s Cross and to AlphaGo’s famous “Move 37.” It is a fairly on-the-nose AI tribute, though in a way that feels deliberate rather than clumsy - largely because DeepMind’s history is woven into it.
More interesting, perhaps, is what comes with the building: a public-facing space called The AI Exchange. So this is not merely branding and architecture theatre. Google appears to want a literal civic venue for AI education and public engagement, which is thoughtful, strategic, or both.
🛒 Webflow buys AI content-generation platform Vidoso to bolster its marketing suite ↗
Webflow is acquiring Vidoso to strengthen its AI-powered marketing stack, adding content-generation capabilities to a platform already centred on building and running websites. Not the loudest AI deal of the day, no, but still the sort of move that shows where the market is drifting.
This is the quieter side of the AI wave - less frontier-model drama, more software companies picking up tools that can automate creative and campaign work. A little less sci-fi, a little more “how do we ship more content with fewer headaches?”
FAQ
Why did Meta delay its Avocado AI model again?
Meta’s Avocado rollout appears to have slipped because internal performance was not yet strong enough. That matters because Meta is trying to compete with OpenAI, Google, and Anthropic at the highest level. The article suggests the delay is not about losing interest, but about failing to meet the standard needed for a major release.
What does Meta reportedly considering Google Gemini technology tell us?
It suggests Meta may have explored using outside technology as a temporary bridge for some products. In context, that would be a notable admission that building a frontier AI model and deploying one reliably at scale are separate challenges. The article presents this as a sign of pressure, not a change in Meta’s broader AI ambitions.
Why is Anthropic’s Pentagon dispute such a big AI news story?
Because the disagreement is not just legal or procedural. The article says Anthropic is fighting a designation that could restrict certain military-related uses of its technology and affect contracts, partnerships, and expected revenue. That turns the issue into a broader test of how AI companies work with national security institutions while trying to preserve their own policy boundaries.
How do the new ChatGPT math and science features change learning?
The update adds dynamic visual explanations for more than 70 core math and science concepts. Instead of only giving text answers, ChatGPT can now show formulas and relationships changing in real time. The article frames this as a shift toward a more interactive learning experience, making ChatGPT feel more like a teaching tool than a static reference.
What is Google trying to do with Ask Maps and immersive navigation?
Google appears to be making Maps more conversational and more useful for planning in one place. Ask Maps is designed to answer complex real-world questions and turn suggestions into map-based decisions without forcing users to jump between apps. The upgraded navigation experience adds clearer route guidance and more immersive visuals to support that broader goal.
What does Webflow buying Vidoso say about where AI is heading?
It points to a quieter but important trend in AI: practical software integration. Rather than focusing on frontier-model headlines, the deal shows how companies are using acquisitions to add content-generation tools into existing products. In this case, the article suggests Webflow wants to make website and marketing workflows more automated, faster, and easier to manage.